IS SHOPPING AT WALMART AN INFERIOR GOOD? EVIDENCE FROM 1997-2010 Mandie R.Weinandt and Michael A. Allgrunn* * Mandie R. Weinandt: Instructor, University of South Dakota, Vermillion, SD 57069. Phone 1-605-677-5690, Fax 1-605-677-5058, E-mail: [email protected] Michael A. Allgrunn: Associate Professor, University of South Dakota, Vermillion, SD 57069. Phone 1-605-677- 8834, Fax 1-605-677-5058, E-mail: [email protected] We test the relative income elasticity of shopping at Walmart and Target using quarterly data from 1997-2010. We seek to isolate the effects of income changes by controlling for price level, retail space, and measures of time. In contrast to Basker (2011), we find that the income elasticity of Walmart shopping , while lower than Target’s, is positive, indicating that shopping at both stores is normal rather than inferior. (JEL D12, L81) I. INTRODUCTION Walmart is often offered up as an example of a company that performs particularly well during recessions. The common narrative is that Walmart offers a low-price shopping experience that consumers value more during a recession than they do when their incomes are higher1. This would seem to be a textbook example of what we economists call an inferior good. A good or service is “inferior” in the economic sense if consumers buy more of it when their incomes fall, other things equal. Put another way, a good or service is inferior if its income elasticity of demand is less than zero. Note that this is different than simply analyzing financial performance during recessions. It would not be enough, for example, to note that Walmart’s earnings rise when incomes fall, as earnings could rise for many reasons. The ideal test would hold prices and supply factors constant so as to isolate the effect of income on demand. In this paper we construct such a test to determine the income elasticity of demand for shopping at Walmart and its close competitor, Target. 1 For example, “Wal-Mart flourishes as Economy Turns Sour” Bustillo and Zimmerman, Wall Street Journal, November 2008 and “McDonalds, Walmart Beat Market Gloom” Andrzej Zwaniecki, IIP Digital, December 2008. II. LITERATURE REVIEW There are a number of studies which examine income elasticity of individual goods. Ito, Peterson, and Grant (1989) attempt to determine the income elasticity of rice in Asian countries. They compare percent changes in real GDP per capita to the percent changes in rice consumption from 1971 to 1985 in fourteen different Asian countries. They found negative income elasticity for rice in economically advanced Asian countries and positive income elasticity for rice in less advanced countries holding own price and substitute prices constant. They suggest that rice becomes an inferior good as the living standards of Asian countries rise. Garrett and Coughlin (2009) examine income elasticity for lottery tickets using county-level panel data for three states in order to determine the relationship between income elasticity and tax-burden. They found that regressively of lottery sales varied both over time and relative to income levels in different states. Studies which examine income elasticity for aggregated goods are less common. Freedman (2003) looks at changes in health care expenditures over time and compares them to changes in disposable personal income to determine income elasticity for health care. Using state level data to determine the relationship between disposable personal income and health care expenditures, they find that health care has positive income elasticity, implying that health care is a normal good. Lu, Thompson, and Tu (2010) analyzed the differences in income elasticities of computers and packaged software with respect to governments, businesses, and individual consumers. They found that computers and packaged software were inferior goods to government agencies, necessary goods for firms, and luxury goods to households. Our study has much in common with Basker (2011), who also sought to compare income elasticity for Walmart and Target. Using data from 1997-2006, Basker used the natural log of the real aggregate wage income as reported by QCEW and the natural log of real quarterly revenues per store for Target and Walmart as the measure of changes in consumption. We perform a similar test to estimate the income elasticities for shopping at Walmart and Target with several differences. First, we have identified the need to control for changes in the relative sizes of each the stores. Without this control, expansions of retail space due to building larger stores during a recession could be mistaken for income inferiority. For example, Walmart and Target have both introduced superstore versions of their previous retail outlets. If a regular store is converted to a supercenter, the number of stores does not change but retail space increases. Second, we exclude revenues from store credit cards from Target’s quarterly revenue data, as they are not part of purchases, but rather are the proceeds from interest charges and fees. Finally, we extend the timeframe of the study to include the most recent recession. III. DATA AND METHODOLOGY We use quarterly sales revenue data from the first quarter of 1997 through the first quarter of 2010, giving us 53 quarters of data for both companies. Since Walmart and Target sell a variety of goods, quantity demanded cannot be defined in the typical way as the number of units purchased. Instead, we use real quarterly revenues, measured in 2010 dollars. Since sales could increase (decrease) due to an increase (decrease) in either the number or size of stores, we use the percentage change in revenue per square foot. This way, we are measuring the effect of changes in consumer demand rather than changes in the scale of the company. All information about Walmart at Target’s revenues and square footage2 was taken from their quarterly and annual filings with the Securities and Exchange Commission. We use two different measures of income: the percentage change in quarterly real GDP per capita; and disposable income per capita, both from the Bureau of Economic Analysis. We also include a regression using QCEW to recreate Basker’s work3. To obtain the ceteris paribus effect of income, we also include several controls. The inclusion quarterly indicators to account for retail sales patterns is straightforward. Controlling for price, however, is not, as both Walmart and Target sell a wide variety of goods and services with presumably autonomous price changes. Since we are looking at goods and services in aggregate, we use the Consumer Price Index (Bureau of Labor Statistics) to adjust all dollar figures for inflation. We also include a time trend control. Table 1 provides summary statistics for each of these variables. 2 Prior to 2004, Walmart did not report their square footage on a quarterly basis but only on an annual basis. They did however; report their stores by store type quarterly. To obtain square footage estimates an average square footage by store type for each year was applied to the number of stores in each quarter missing square footage data. For example, in January 2003, Walmart had 1,258 Supercenters with an average of 186,495.23 square feet. This average was multiplied by the number of Supercenters in the three preceding quarters to obtain the number of total square feet in Supercenters for that quarter. The same was done with regular Discount Centers and Walmart’s, more recent, Neighborhood Markets to obtain a total count on Walmart’s square feet per quarter prior to 2004 when actual data was available by quarter. When applying this methodology to quarters with actual square foot data, we find that the difference between the simulated square foot information and actual square foot information does not exceed 2%. Target reports actual information quarterly. 3 The authors would like to thank Basker for providing her original data and do-file. TABLE 1 Summary Statistics Table 1. Summary Statistics Std. Variable Description Mean Min Max Dev. Bureau of Economic Analysis Consumer Price Index Consumer Price Index. Quarterly 187.6 18.8 159.9 218.47 price index for all consumer goods. Walmart: Revenue Per Quarterly Revenue per Store in 14.19 3.11 7.88 19.58 Store Millions Walmart: Revenue Per Quarterly Revenue per Square Foot 101.34 11.48 73.48 121.27 Square Foot Target: Revenue Per Quarterly Revenue per Store in 8.46 1.76 5.66 12.95 Store Millions Target: Revenue Per Quarterly Revenue per Square Foot 69 12.01 52.29 100.33 Square Foot Quarterly Census of Bureau of Labor Statistics quarterly 1605.8 Employment and count of employment and wages 1252.5 197.45 888.91 5 Wages reported by employers in billions. Bureau of Economic Analysis GDP quarterly Gross Domestic Product 39429 5749.13 29947 47666 Per Capita Bureau of Economic Analysis 29330. Disposable Income quarterly Disposable Income Per 4493.04 21932 36022 2 Capita n=53 Figure 1 shows our dependent variable over time. Not surprisingly, both Walmart and Target show significant seasonality in revenue changes. We also see much more variation in percentage change in revenue per square foot for Target. FIGURE 1 Percentage Change in Revenue per Square Foot, 1999-2010 0.6 0.5 in 0.4 0.3 0.2 Change 0.1 0 1 6 36 16 21 26 31 41 46 51 -0.1 11 -0.2 Percentage Percentage -0.3 Revenue Per Square Foot Square Revenue Per -0.4 Time Period Walmart Target The equations to be estimated take the following form: (1) where t denotes quarters, Walmart is a dummy variable, ln(income) is the natural log of the income measure, ln(income)•Walmart is the interaction between income and Walmart, time is a simple time trend, (time)•Walmart is the interaction between time and Walmart, Q2, Q3, and Q4 are quarter indicators, and u is the error term.